capital gains tax rates and stock market volume

National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
CAPITAL
GAINS TAX RATES
AND STOCK MARKET
VOLUME**
YOLANDA K. HENDERSON*
ABSTRACT
This study examines
the relationship
between aggregate stock market volunw and
changes in capital gains tax rates. The tax
cuts of 1978 and 1981 usually were founa'
to
to increase trading. The tests failed
detect any significant decrease in volume due
to the increase in the capital gains tax rate
in 1987, however. Accordingly, if revenues
and tax rates were inversely related around
the time of the Tax Reform Act of 1986, as
some have claimed, this relationship might
be due to a changing mix of sales among
stocks with gains and losses or to changing sales volume for other assets. These
mechanisms deserve further study, as do
non-tax motivations for selling stocks.
1. Introduction
REVIOUS time-series
analyses
are
Pinconclusive
about the revenue yield
from changing the capital gains tax rate.
The estimates have varied for at least
three reasons. First, the effect of tax policy on capital gains realizations
is very
sensitive to the choice of specification.
Second, the revenue effect from a given
realization
response
depends
on the
method used to construct average effective tax rates with and without the tax
change. Third, the relationship
between
realized capital gains and the rate of tax
is unstable because it depends on factors
that vary over time, such as the size of
accrued gains and the anticipation
of future tax rates. Among the recent studies
illustrating
these points are Auerbach
(1988), Henderson (1989), Kiefer (1990),
Toder (1988), and U.S. Congressional
Budget Office (1988).
Many observers would expect, however,
that stock market volume responds
slgnificantly to the capital gains tax rate.
Kiefer (1990) developed an analytical
inodel with this result, and Slemrod (1982)
found empirical confirmation
in his analYsis of the Revenue Act of 1978. One rea*FederalReserveBankofBoston,Boston,MA02106.
411
son for the responsiveness
of stock mar@gt volume would beLtb$ -_tk e c4WiU_a gains
t ax is-a- -m'ajo -r-e-Ie-me-@nto-f the transactions
costs associated
with selling stock. Another reason would be that ownership of
corporate shares is concentrated
among
the very highest income taxpayers,
who
typically tend to respond the most to tax
incentives of all sorts. Responsiveness
of
corporate share trading would add credence to the conclusion that lowering this
tax might cause a minimal revenue loss,
and possibly a revenue gain.
This study supplements
previous evidence on the effects of capital gains policy
on stock market volume. Reductions
in
capital gains tax rates in 1978 and 1981
are usually found to raise volume. However, the increased tax rates under the Tax
Reform Act of 1986 do not appear to have
reduced volume significantly.
This study does not attempt to characterize in full the complex relationship
between capital gains taxation and tax revenues. For example, it does not encompass
assets apart from corporate stock, such as
real estate or other business property. Nor
does it consider the revenue impact of
other sources of income (such as dividends) that may be influenced by capital
gains tax policy. Furthermore,
revenue
effects from sales of corporate stock depend not just on volume, but also on the
pattern of accrued gains and losses and
on the tax status of the sellers. Nevertheless, because the issue is so complex, information about each component of the
link between rates and revenues is valuable. Because of the availability
of recent
data, stock market volume is a particularly good source of evidence on the Tax
Reform Act of 1986.
The next section briefly reviews the
theory of how taxes and other factors influence stock market volume. Section III
summarizes
previous empirical studies of
stock market volume and updates the
Slemrod (1982) study to consider the Eco,omic Reovery Tax Act of 1981 and the
Tax Reform Act of 1986. Because the ex-
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
412
NATIONAL
TAX JOURNAL
tension of the sample period exposes some
weaknesses in specification, Section IV
develops new specifications and Section V
presents results of these revised estimates. Section VI interprets the findings
and offers suggestions
for future research.
II. The Theory
Volume
of Stock Market
Tax Effects
The capital gains tax tends to reduce
stock market volume because it creates a
transactions cost. Without a capital gains
tax (and in the absence of risk considerations), investors would trade shares on
which they have accrued a gain if the expected future rate of return on alternative stock exceeds the expected future rate
of return on their existing holdings. With
a capital gains tax, the expected rate of
return on their alternative investment
must be higher, in order to offset the tax
paid on the appreciation in the existing
shares when they are sold. The higher this
alternative expected return, the less likely
investors are to find a new uivestment that
satisfies the conditions for selling.' However, for stocks with accumulated losses,
the capital gains tax provides an incentive to sell shares so that these losses may
be used currently to lower taxable income. 2 The unlocking effect of a reduction
in the capital gains rate thus depends on
the size of accrued gains and on the distribution of investment opportunities.
Because individuals pay capital gains
taxes only as they file annual returns, tax
policy also influences the timing of gains
and losses throughout the year. All else
equal, sales volume for stocks with accrued gains should be abnormally low toward the end of the year because postponing their sale by a month or two
postpones payment of capital gains tax by
a full year. Conversely, sales volume for
stocks with losses should be abnormally
high toward the end of the year because
investors will want to apply these losses
against their taxable income as soon as
possible.
[Vol. XLJII
Tax effects should be more important
determinants
of aggregate volume the
larger the participation
of individual
investors. Among institutional
investors,
pension funds and university endowment
Rmds, for example, are totally exempt from
capital gains taxation. Finally, tax effects
on volume will be less important if individual investors routinely evade taxes by
underreporting
capital gains.'
Non-tax Variables Influencing
Market Volume
Stock
The theory of stock market volume apart
from tax effects centers around the role of
information. Heterogeneity
of investor
characteristics
or opinions is deemed necessary for trade to take place. For example,
Varian (1987) outlines the following
,
no trade" theorem: if it is generally believed that traders are risk averse, rational, and have the same prior expectations about stock prices, then there will
be no trades. If one investor has information that makes him willing to trade
at current prices, others will be unwilling
to trade because they will assume that he
has superior information. Karpoff (1986)
modifiesthis
theorem to allow for some
trading if investors have heterogeneous
liquidity or speculative desires. In both
these models, the arrival of new information can increase trading if different
investors interpret it differently. In Karpoits model, new information can also
generate trading if investors begin with
diverse prior expectations.
111. Empirical
Results
Comparisons
of Trading for Stocks with
- ns
and
Stocks
with Losses
L'al
Most existing studies of stock market
volume have attempted to explain differences in volume between stocks with accrued gains and those with accrued losses,
rather than aggregate volume levels over
time. Using a sample of stocks listed on
the New York Stock Exchange, Dyl (1977)
found evidence of abnormally low volume
in December for those with accumulated
gains and abnormally high volume for
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41
CAPITAL
GAINS TAX RATES
AND STOCK MARKET
VOLUME
413
those with accumulated
losses, as expected from tax incentives. But he found
insignificant differences in January. Lakonishek and Smidt (1986) found that the
average turnover rate for stocks whose
prices have increased is much higher than
that for stocks whose prices-hiivectecreased, and suggested that this might be
due to considerations
of risk. That is,
investors wish to sell stocks whose prices
have appreciated in order to diversify their
portfolios. But they also found that the
pattern of trades in December and January supports the hypothesis of tax-motivated trading.
Ferris, Haugen,
and
Makhija (1988) reached similar conclusions about relative trading volumes for
stocks with accumulated
gains and losses,
and interpreted
it as evidence in favor of
analyzed aggregate stock market volume,
and found that the lower tax rates introduced in the Revenue Act of 1978 strongly
increased sales. Because Slemro&s study
forms the starting point for the empirical
work in this study, its specifications
and
fitidings are set ott hfare-tg-some detifil.
Slemrod estimated
separate
monthly
aggregate time-series regressions
through
May 1981 for the New York Stock Exchange, the over-the-counter
market, and
the American Stock Exchange. (May 1981
was a natural ending date because the
Economic Recovery Act introduced
a further reduction in capital gains tax rates
in January
1981.) Re hypothesized
that,
in the absence of the 1978 tax reform,
trading volume would depend on changes
in information
about the state of the
the
economy,
since
this would
cause
owners
to revise their optimal
portfolios.
As measures of news on the economy,
he used the
absolute
change
in monthly
stock market
prices, the volatility
of stock prices within
the month,
and the absolute
month-tomonth change
in three macroeconomic
indicators-the
six-month
Treasury
bill rate,
the unemployment
rate, and the inflation
"disposition
effect."
The
disposition
effect is premised
on a utility function
that
is concave
over gains
and
convex
over
losses. An investor
holding
a stock with
accrued
gains and facing
equal
probabilities of future
gains and losses will have
higher utility
from realizing
the gain than
from the expected
payoff from holding
onto
the shares. Furthermore,
the authors
noted
that
abnormally
high
volume
in stocks
with
accumulated
gains
consistent
with both the
in January
is
tax hypothesis
and the disposition
hypothesis.
Bolster,
Lindsey,
and Mitrusi
(1989)
looked specifically
at how anticipation
of
higher
capital
gains
taxes
in 1987 affected
typical
volume
patterns
following
the passage
of the Tax Reform
Act in October 1986. They hypothesized
that the Act
reduced
the tendency
to postpone
realization of capital
gains in late 1986 because
the increase
in the capital
gains
rate in
1987 was a strong
offset to the usual advantage
of postponing
payment
of tax.
Similarly,
the increase
in rates
should
have lowered
the tendency
to realize
capital losses. They found support
for the first
hypothesis,
but not for the second.
The Slemrod
Study
Trading
Volume
ofaggregate
Unlike
these
studies
dealing
ative selling
patterns
for stocks
and
stocks
with
losses,
with relwith gains
Slemrod
(1982)
rate based
Slemrod
resent
the
form
Act
changes
in
on the consumer
price index.
used dummy
variables
to repperiod
following
the Tax Reof 1978. This
act introduced
two steps. As of October
31,
1978, the inclusion
rate for capital
gains
was lowered
from 50 percent
to 40 percent. Then in 1979, the structures
of the
minimum
tax and maximum
tax (which
affected
individuals
in the top brackets
and
those with high capital
gains income) were
changed
to provide
more favorable
treatment for capital
gains.
Slemrod
therefore
tried separate
dummy
variables,
one for
November
and December
1978 and one
starting
in 1979. One specification
of the
latter
dummy
hypothesized
that the tax
reform would have a constant
effect on the
level of stock market
volume.
The other
specification
involved
a quadratic
function, so that the effect of tax reform would
build over time,
and then eventually
decline.
Slemrod's
results
indicate
a sizable
effect on stock market
volume
starting
in
1979. In the constant-effects
version
of the
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
414
NATIONAL
TAX JOURNAL
specification (shown in Table 1), during
January 1979 to May 1981 average daily
volume on the NYSE was estimated to be
ESTIMATED EQUATIONS
1978 Law Dummy,
Nov-Dec 1978
1978 Law Dummy,
Starting
January
1979
13.4 million shares, or 50 percent, higher
than it would have been in the absence of
tax reform. Volume on the OTC was 10.7
TABLE 1
FOR AVERAGE DAILY VOLUME: SLEMROD STUDY AND EXTENDED
SAMPLE'
New York Stock
Exchange
Independent
Variable
Constant
[Vol. Y.LIII
Over-the-Counter
Market
American
Stock Exchange
1965:2 to
198liS
1965:2 to
1990:6
-2.00
(2.72)
1972:2 to
198liS
1972:2 to
1990:3
-13.02xx
(4.32)
-.001
(3.24)
-3.05
(10.65)
-2.16
(3.80)
-16.10
(11.00)
.740
(.92)
-.031
(1.37)
10.74**
(1.89)
-4.61
(4.85)
3.02**
(.41)
2.32**
(.57)
13.42**
(1.25)
13.52**
(3.97)
1969:2 to
198li5
1969:2 to
1990:6
3.30**
(.44)
1981 Law Dummy,
Starting June 1981
38.§7**
(3.60)
18.58**
(4.39)
1.84**
(.49)
1986 Transition
Sep-Dec 1986
46.65**
(7.65)
26.70**
(7.93)
2.99**
(.98)
66.05xx
(3.34)
45.07xx
(4.47)
4.63xx
(.47)
Dummy,
1986 Law Dummy,
Starting
January
1987
Absolute Percent Change
in Stock Price Index
1.30**
(0.32)
-.118
(.14)
.428
(.33)
.056*
(.03)
.074**
(.03)
-1.21
(.79)
3.88**
(1.95)
.616
(1.11)
3.70
(2.40)
-.588xx
(.23)
-.021
(.26)
.213
(.64)
-8.05xx
(2.16)
.331
(.81)
-6.51xx
(2.30)
-.010
(.19)
-.685xx
(.28)
Absolute Change
in Inflation
Rate
-1.72
(1.57)
-3.80
(4.16)
-.627
(2.03)
2.63
(4.82)
-.535
(.46)
-.055
(.56)
Absolute Change
in Unemployment Rate
.291
(2.10)
-11.55X
(6.56)
-4.17
(3.31)
-11.64
(7.38)
-1.12x
(.75)
-1.63x
(.88)
Intra-Month Stock
Price Volatility,
Standard & Poor's
Index
Absolute Change
in Six-Month
Interest Rate
Time Trend
R Al;u'sSt!dtiforicBegrees
S2't
of Freedom
Root Mean Squared
Mean of Dependent
Variable
Durbin-Watson
Error
.277**
(.13)
.117**
(.008)
.161**
(.024)
.079**
(.025)
.423**
(.057)
-.012**
(.004)
-.002
(.005)
.841
.936
.649
.911
.404
.762
____b
14.64
__b
55.83
.642
.613
-b
.339
14.90
-b
1.87
50.25
-b
6.17
.323
.652
.567
a DeWgont variable
is measured in millions of shares.
The numbers in parentheses are
stan@ird errors.
bnot reported
*Theoretically * acceptable sign and significantly
different
from zero at the 10% level.
**Theoretically
acceptable sign and significantly
different
from zero at the 5% level.
xtheoretically
unacceptable sign and significantly
different
from zero at the 10% level.
xxtheoretically
unacceptable sign and significantly
different
from zero at the 5% level.
Source: Slemrod (1982) for sample through 1981:5 and author's estimates for sample through
1990:6.
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41 CAPITAL GAINS TAX RATES AND STOCK MARKET VOLUME
million shares (91 percent) higher, and
volume on the AMEX was 3.0 million
shares (119 percent) higher. Over this
same interval, the nonlinear specification
shows an average volume increase of 14.7
million shares on the NYSE, 14.9 million
on the OTC, and 3.4 million on the AMEX.
Given the U.S. Treagury-(1985@ egbnfflte
that the average tax rate on capital gains
fell by about 9 percent (from 18.2 percent
to 16.5 percent), Slemro&s measured increases in volume imply an increase in
revenues from the capital gains tax on
corporate stock over this period.
In contrast to the strong performance of
the tax variables, the informational variables added little to the explanatory power
of the equation as their coefficients were
estimated with high standard errors. The
only other strongly significant variable
was a time trend.
Updated Results
This study first re-fits Slemrod's equation to a longer time period. Three new
tax dummies are introduced to capture the
effects of the Economic Recovery Tax Act
of 1981, the transition to the Tax Reform
Act of 1986, and the period during which
the Tax Reform Act became effective.
Through a series of rate reductions, the
1981 Act generally lowered marginal tax
rates by 23 percent, and the top-bracket
rate from 70 percent to 50 percent. Although the inclusion rate for capital gains
was kept at 40 percent, the top capital
gains rate was effectively lowered from 28
percent to 20 percent in June 1981 instead of being delayed until the phase-in
of lower rates. The Act also lowered the
top rate under the alternative minimum
tax from 25 epreent to 20 percent. In the
regressions, the dummy variable for the
1981 Act is set equal to 1 starting in June
1981. The Tax Reform Act of 1986 eliminated the capital gains preference starting in 1987, so that capital gains are now
taxed at the same rates as ordinary income. As a result, the top-bracket rate on
capital gains increased from 20 percent to
28 percent .4 The 1986 law dummy is set
equal to 1 starting in January 1987. The
regressions also include a transition
415
dummy, equal to 1 from September 1986
through December 1986. The Tax Reform
Act was approved by a Congressional conference committee in September 1986, and
was enacted in October 1986, but the capital gains provisions were not effective
until the following January. This delay
prldvidgt I tivdttbrer an ol)porttiiilly to realize gains while the old, lower rates were
still in effect.
The updated regressions use tax dummies entered in level form. Slemrod's
quadratic formulation was not used with
this extended sample because it implies
that a lower capital gains tax eventually
lowers stock market volume, contrary to
the predictions of theory. His sample period following the 1978 reform was too
short to encompass this period of lower
volume, so this problem with the dummy
variable did not affect his results. In the
context of the longer sample period, however, it is inappropriate
to use this formulation.
During Slemro&s sample period, stock
market volume was fairly stagnant until
the 1978 reform. Since that time, the
overall trend rate of increase has been
much greater. However, the patterns do
not clearly correspond to changes in tax
law. The rate reduction in 1981 did not
coincide with an inunediate rise in trading. Trading did pick up around the end
of 1982, but stopped increasing early in
1986, well before a change in capital gains
policy appeared likely. Volume was higher
in early 1987, after the rate increase, than
in late 1986. (These data are sunnnarized
for the NYSE in Appendix Table 1.)
Table 1 indicates the results with the
sample extended through early 1990. For
the 1987 law, the equations continue to
show strong positive effects equal to 13.5
million (51 percent) on the New York Exchange and 2.3 million (74 percent) on the
American Exchange. But the equations
indicate an insignificant negative coefficient for the over-the-counter
market. The
1981 change produced positive volume
changes in all three markets. Measured
relative to volume over the two-year period following the enactment of the legislation, the increases were 155 percent
on the NYSE, 95 percent on the OTC, and
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
416
NATIONAL
[Vol. XLIII
TAX JOURNAL
45 percent on the AMEX. These kinds of
changes imply positive revenue effects
from stock sales over this period because
average tax rates declined by about 13
percent (to 14.3 percent; U.S. Treasury
Department 1985). The equations also indicate a positive volume response in the
late 1986 transition period: 47 percent on
the NYSE, 34 percent on the OTC, and 39
percent on the AMEX. The equations fail
to find the expected negative trading effect starting in 1987: all the coefficients
for this variable are positive and significantly different from zero.
In addition to the unexpected results for
the 1986 Act in all markets and for the
1987 Act in the over-the-counter
market,
another problem is the negative constant
term in the equation for the over-thecounter market. This problem probably
arises because the trend in volume is
sharply nonlinear in the extended sample
period. The findings suggest that the
original specification is inappropriate for
this longer time interval.
IV. Revised Specifications
This section discusses variations on the
Slemrod equation with the longer sample
period. The revisions include adjustments
for the number of shares outstanding and
for institutional
activity, respecification
of the economic surprise variables, new
tests of the role of differences of opinion
and turn-of-the-year
tax effects, a dummy
for the October 1987 stock market crash,
and correction for serial correlation.
Adjustment for Shares Outstanding
Over time, the number of shares outstanding has changed with changes in the
number of companies listed on the exchanges, new issues of shares, repurchases of shares, and stock splits. A specification in which the dependent variable
is the number -6f -shares tra-ded may produce heteroscedastic
errors because the
dependent variable has a sharp upward
trend but the independent variables generally do not. Also, it is unlikely that a
change in the capital gains rate would
have a level effect on trading if the number of shares outstanding changes. This is
especially so in the case of stock splits because the same dollar adjustment in portfolios requires trading more shares. For
these reasons, the adjusted equations use
as the dependent variable the turnover
rate-that
is, sales volume as a percent
of shares outstanding at the beginning of
the month.
Because data on shares outstanding are
not available as far back as volume data,
using this turnover rate necessitates
shortening the sample period and interpolating from annual data to create
monthly data in the early years. Figure 1
indicates that turnover rates have increased over time. However, unlike volume which has risen exponentially, turnover rates have risen linearly with the
exception of NYSE trades.
Adjustment
for Institutional
Activity
The effect of tax policy changes on trading volume is unlikely to remain constant
if the mix of participants
changes between taxable and nontaxable investors.
According to estimates of "retail" trading
by the Securities Industry Association
(SIA), since 1985 noninstitutional
investors have accounted for just one-fifth to
one-third of the shares traded on the
NYSE, one-half on the AMEX, and fourfifths on the OTC (Figure 1). (The SIA
"adjusted retail" estimates, available only
since 1987, attribute an even higher share
of NYSE trades to program trading and
other professional trading strategies.)
These data are not available for previous
years, but ownership data suggest that the
participation of individual investors has
fallen over the past two decades. According to the Federal Reserve Flow of Funds
acounts, households owned about 80 percent of corporate equities in the early
1970s, and only 60 percent in the late
1980s. 5
To approximate NYSE trading volume
by noninstitutional
investors, this study
subtracts out shares sold in large blocks
(of at least 10,000 shares), since these are
typically put together on behalf of insti-
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41
CAPITAL GAINS- TAX RATES AND STOCK MARKET VOLUME
417
FIGURE 1
AVERAGE DAILY TURNOVER RATE'
New York Stock
Exchange
percent of shares outstanding
0.35
0.3
Total
0.25
0.2
Excluding
ot!W
Large Blocks
0.15
1
0.05
0, L
68 69 70 7172
twi
73 74 75 76 77 78 79 80 8182
Over the Counter
............
83 84 85 86 87 88 89 90
Market
percer7t of shams outstanding
0.5
0.4
Total
0.3
NoninstrbAonal
Sales
Excludng
Large Blocks
0.2
0.1
0
j@
68 69 70 71 72 73.@j";15
74
76 7L77879808
American
Stock
.@ji,@2L.@L3 8"4@@8L5'@6
L 87 88 89 90
Exchange
percent of shares outstanding
0.25
Totad
tad
0.2
V tA
0.15
0.1
FReVtaii
0.05
0
.........
68 69 70 71 72 73 74 75 76 77 78'@;"80"8
a Quarterly data
sources. see append.
. . .
9 90
82 8"3"8"4't5 [email protected]
7
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
418
NATIONAL
TAX JOURNAL
tutional clients. (Slemrod had in fact suggested such an adjustment.) These data
are available starting in 1968. Similar
data are available for the OTC, starting
in 1983. We smoothed both sets of data
down to zero in earlier years to permit a
longer estimation period. Figure 1 indicates that this adjustment brings the level
of trading activity down closer to the direct estimates of noninstitutional
activity
in recent years.
Changed Specification
Surprise Variables
of Economic
This study makes four revisions to the
specification of economic surprise variables. First, the intra-month volatility
measure was calculated separately for the
over-the-counter
market, based on the
National Association of Securities Dealers (NASDAQ) composite. Second, the inflation and unemployment rate terms were
lagged by one month, so that their effect
on volume corresponds to their date of announcement. Third, the expected inflation rate and unemployment
rate are
modeled as outcomes of ARIMA processes.' Slemrod's model assumes that
market participants
expect current values of these variables to continue, so that
any change is a surprise. In this revision,
the unexpected change in the interest rate
was omitted because it was highly correlated with unexpected changes in inflation and unemployment.
forecast. These surveys are conducted
quarterly, so that in this version of the
specification, the regressions were estimated using quarterly
rather than
monthly data.
The End-of-Year Effect
The year-end theory of tax behavior indicates unusually low volume for stocks
with gains and unusually high volume for
stocks with losses. At the beginning of the
following year, there should accordingly
be unusually high volume for stocks with
gains and unusually low volume for stocks
with losses. The aggregate data do not allow a distinction for these two categories
of shares, but the thrust of the theory was
tested by including two new dummies. For
each December and each January, these
dummy variables were equal to the change
in stock prices over the previous twelve
months. In other months, the dummy
equalled zero. The anticipated sign was
negative for December and positive for
January.
October 1987 Dummy
As indicated in Figure 1, trading volume became unusually high around the
time of the October 1987 stock market
crash. Accordingly, the new specifications
include a dummy variable equal to one for
that month, and zero in all other months.
Alternative
The Role of Differences of Opinion
As discussed above, theories of stock
market volume include a role for differences of opinion. These theories are not
specific as to the exact information about
which this difference arises, but one set
of equations tests this view by introducing as new explanatory variables the dispersion of one-year ahead forecasts for real
GNP growth and inflation in a survey
conducted since 1968 by the National Bureau of Economic Research.7 Dispersion
was measured, alternately, by the standard deviation divided by the mean forecast and by difference between the first
and third quartiles relative to the median
[Vol. XLIII
Lag Structures
As discussed
above, the quadratic
structure for the dummy variable is not
appropriate
for the extended sample.
However, alternative regressions tested
whether the trading effect is different in
the first two or three years after capital
gains rates are changed by introducing a
dummy variable that decays after this interval. (In Slemrod's work with nonconstant dummies, the peak effect on trading
was estimated to occur at 32 months for
the NYSE and 25 months for the AMEX.
For the OTC, it occurred at 225 monthsalmost 19 years-after
the tax change, an
implausible
delay.) These alternative
dununy variables never provided superior
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41 CAPITALGAINS TAX RATESAND STOCKMARKETVOLUME
explanatory
power than the constant
dummies, so these results are not reported.
Correction for Autocorrelation
The revised regressions include a correction for serial correlation of the errors.
This provides more precise estimates of
the coefficients.
V. New Times-Series
Evidence
Table 2 shows the respecified regressions based on the turnover rate, for five
different dependent variables: total New
York Stock Exchange, NYSE excluding
large blocks, total over-the-counter
market, OTC excluding large blocks, and the
American Exchange. Additionally,
the
NYSE equation excluding large blocks is
estimated for a shorter time period, starting in May 1975. At that time, the Securities and Exchange Commission prohibited member brokers from setting fixed
mimmum commissions, a practice that had
increased transactions
costs and caused
some transfer of activity off the New York
Stock Exchange (Campbell 1982, pp. 395396).
In the equations with the longer interval for the New York Stock Exchange, the
1978 and 1981 laws had significant positive effects on turnover, though lower
than estimated in the volume equations.
The measured increases were 27 percent
and 34 percent, respectively, according to
the total turnover specification. When
large blocks were excluded, the estimates
were 25 percent from the 1978 law and 15
percent from the 1981 law.
In the remaining equations, the 1978
law tended to have a positive effect on
turnover, but this was measured with a
large standard error. The 1981 dummy
appeared with a negative sign, but was
also insignificant. These anomalous results may be due to the lack of sufficient
observations prior to these law changes
(as suggested by the differences between
results in column 2 and column 3), and
therefore should not be taken to be as authoritative as the NYSE estimates in the
first two columns.
419
In almost all the equations, the law
change in 1986 failed to produce higher
turnover in late 1986 or lower turnover
starting in 1987. This result is undoubtedly due to the underlying data (Appendix Table 1), which indicate that turnover
rates did not rise in late 1986 or fall in
the following year. (Additional experiments used volume less large blocks as a
function of the independent variables from
Table 2, and turnover as a function of the
independent variables from Table 1. The
1986 transition dunnny sometimes had a
significant positive coefficient, but the
1987-90 dummy invariably also had a
positive coetticient, in contradiction with
theory.) The concluding section provides
possible interpretations
of these results.
The coefficient for the January dummy
was positive in all the regressions, indicating that shareholders had a tendency
to delay realizing gains when stocks had
appreciated in value. This is consistent
with both the tax hypothesis and the disposition hypothesis. The coefficient for the
December dummy was insignificantly different from zero in almost all cases, perhaps indicating the difficulty of measuring the extent of losses on an aggregated
basis.
The monthly absolute change in stock
prices and the intra-month volatility of
stock prices, both measures of economic
news, each generally had a positive effect
on turnover. Slemrod had found less significant effects. Unexpected changes in
inflation and unemployment
tended to
raise turnover, but usually these effects
were not significantly different from zero.
The stock market crash dummy had a
positive coefficient that was usually statistically significant. The revised estimates continued to indicate a significant
positive time trend in most cases. The autoregressive
parameter
was negative,
which indicates that-all else equal-bigh
volume in one month tends to be offset by
low volume the next month.
Finally, quarterly regressions
(not
shown) used a different informational
variable, the dispersion in economic forecasts of output and inflation. The coefficients for these variables tended to be
positive and sometimes were significantly
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
ESTIMATED
Independent
Variable
EQUATIONS
New York
Stock Exchange
1968:2 to 1990:6
FOR AVERAGE
NYSE Less
Large Blocks
1968:2 to 1990:6
TABLE 2
DAILY TURNOVER
DUMMIESA
NYSE Less
Large Blocks
1975:5 to 1990*6
Constant
.0635**
(.012)
.0648**
(.006)
.0267*
(.015)
1978 Law Dummy,
Nov-Dec 1978
.0082
(.016)
.0051
(.010)
-.0021
(.011)
.0272*
(.016)
.0185**
(.009)
.0062
(.008)
1981 Law Dummy,
Starting
June 1981
.0418**
(.014)
.0125*
(.007)
.0006
(.008)
1986 Transition
Sep-Dec 1986
.0191
(.015)
.0132
(.010)
.0065
(.010)
.0328xx
(.015)
.0094
(.008)
-.0083
(010)
.0975**
(.016)
.0325**
(.011)
.0139
(.013)
-.0001
(.0002)
-.0001
(.0001)
-.0001
(.0002)
.0006**
(.0002)
.0004**
(.0002)
.0004**
(.0002)
Absolute Percent CFonge
in Stock Price Indii
.0018**
(.0003)
.0012*1
(.0002)
.0013**
(.0003)
Intra-Month
Stock
Price Volatility
.0054*
(.003)
.0042**
(.002)
.0102**
(.003)
1978 Law Dummy,
Starting
January
1979
Dummy,
1986 Law Dummy,
Starting
January
October
Dummy
December
January
1987
1987 Stock Market
Dummy
Dummy
RATE,
US
Over1976:
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
Table
Unexpected
Inflation
Period
Change in
Rate, Lagged One
Unexpected
Change in
Unemployment Rate, Lagged
Period
Time Trend
Autoregressive
Sumary
Parameter
One
-.0009
(.004)
.0013
(.003)
.0037
(.004)
-.0004
(.007)
.0013
(.005)
.0071
(.007)
.0002
(.0001)
-.0000
(.0001)
.0003**
(.0001)
-.7650**
(.040)
-.6762**
(.046)
-.6408**
(.065)
Statistic
R2
.935
.785
.771
.0963
Mean of Dependent
Variable
.1413
.0883
Root Mean Squared
Error
.018
.012
adependent
variable
is volume as a percent of shares outstanding
at the
*Theoretically
acceptable
sign and significantly
different
from zero at
**Theoreticaly
acceptable
sign and significantly
different
from zero at
xtheoretically
unacceptable
sign and significantly
diffrent
from zero at
xTheoretically
unacceptable
sign and significantly
different
from zero
: ource:
2 Continued
Author's
estimates
described
in the text.
.012
beginning of the period.
the 10 percent level.
the 5 percent level.
the 10 percent level.
at the 5 percent level.
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
422
NATIONAL
TAX JOURNAL
different from zero, indicating that they
are a somewhat useful approach to measunng differences of opinion.
VI. Summary and Conclusions
This paper has investigated whether
stock market volume, as measured by aggregate data from the New York and
American Stock Exchanges and the overthe-counter market, expanded following
cuts in capital gains tax rates and contracted following the recent increase in
capital gains tax rates. The methodology
is based on the study of Slemrod (1982),
who found very large responses to the
Revenue Act of 1978. The current study
extends the sample period to June 1990
and makes several changes to-Slemrod's
specification.
The results cast doubt on whether
changes in capital gains tax rates reliably
have an inverse effect on revenues through
responses in stock market volume. In the
preferred specification, which uses turnover rates as the dependent variable, the
tax cuts of 1978 and 1981 were found to
increase trading volume on the New York
Stock Exchange. Results for the other exchanges were positive when volume was
used as the dependent variable, but not in
the turnover rate specification. This difference may be due to the shorter time
period necessitated by lack of early data
on shares outstanding, however. No specification indicated a decrease in volume
due to the increase in the capital gains
tax rate in 1987. Volume was abnormally
high in January if stocks had appreciated
in value. Economic news, as measured by
changes in stock prices, raised trading
volume, and differences in projections for
the economy were sometimes a significant factor. These results are consistent
with the raw data for the 1986 to 1990
period. Turnover rates did not rise in late
1986 or fall in early 1987, and their pattern was somewhat correlated witly'stock
price volatility.
The insignificance of the post-1986 tax
law dummy does not rule out the POB$Ibility that capital gains tax rates may affect revenues through selective sales of
[Vol. XLIII
stocks with very high accrued losses when
rates are relatively high rather than
through a general decrease in volume.
Analytically, the incentive to sell stock
when rates are raised is greater for stocks
with high accrued losses. Future empirical research might therefore continue to
investigate to what extent tax policy influences the distribution of realized gains
and losses. Also, swings in revenues found
in some other studies may reflect sales of
other assets, such as real estate, not considered in this study.
Future research should also investigate
further the role of non-tax factors influencing capital gains realizations.
Share
repurchases have been very large in recent years, and have been motivated by
factors such as firms' cash flow position
relative to their investment opportunities
and their leverage ratios relative to desired values (Bagwell and Shoven 1988).
Additionally, share acquisitions through
tender offers have had a substantial effect
on realizations in recent years, and affect
revenues without changing volume recorded on the stock exchanges. These
transactions stem from a broad variety of
motives (Auerbach and Reishus 1988).
These factors apart from capital gains
taxes may well have had an important effect on capital gains tax revenues in the
last several years, and may account for
some of the patterns observed around the
time of the Tax Reform Act of 1986.
ENDNOTES
.*rhe views expressed in this paper are solely those
of the author, and do not represent official positions
of the Federal Reserve Bank of Boston or the Federal
Reserve System. This research has benefited from
helpful discussions with Don Fullerton, Patric Hendershott, Joseph Minarik, Larry Ozanne, Joel Slemrod and Eric Toder. Two anonymous referees provided
e=eiient suggesuons that improved the paper greatly.
I am grateful to Natalie Inman and JefFrey B. Liebman for extensive research assistance.
'The appreciotlon on sham held until death escapes tax idtog-ether, thus co@-sti-t@t-ing--anothe@r
reason not to sell.
'Capital losses are deductible in full against capital
gains, and a maximum $3000 of additional capital
losses are deductible against ordinary income. Excess
losses must be carried forward to subsequent years.
For a thorough analysis of the timing option asmi-
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41
CAPITAL GAINS TAX RATES AND STOCK MARKET VOLUME
ated with stock ownersinp and the influence of tax
policy, see Constantinides (1983). Timing was even
more important prior to 1987, when the tax code distingtdahed between short-term and long-term holding
periods. An alternative influence is the so-called disposition effect (Shefrin and Statman (1985) and Ferris, Haugen, and MANA (1988)), which indicates that
indivklusb wfli-hdd onto stocks with accumulated
losses and sell stocks with accumulated gains.
38w Poterba (1987).
4During 1987, the top income tax rate was 38.5 percent, but the top-bracket rate on capital gains income
was reduced immediately to 28 percent. As a result
of the phase-out of personal exemptions and other
preferences, some high-income taxpayers face a rate
of 33 percent.
'These data include the ownership of corporate equities by personal trusts and nonprofit organizations,
which are not provided separately.
'Inflation was modeled as seasonal-nonseasonal
ARIMA (0, 1, 1) (0, 1, 1) 12, and the unemployment
rate was modeled as seasonal-nonseasonal
ARIMA (2,
1, 0) (0, 0, 2) 12. The interest rate model was ARIMA
(0, 2, 2), but surprises in this variable were omitted
from the final specification. I am grateful to Natalie
Inman for estimating these ARIMA models.
7
1 am grateful
to Wayne Gray for providing these
data.
REFERENCES
Auerbach, Alan J. 1988. "Capital Gains Taxation in
the United States: Realizations, Revenue and Rhetoric-A Review." Brookings Papers on Economic
Activity: 2, pp. 595-631.
-,
and David Reishus. 1988. "The Effects of Taxation on the Merger Decision." In Corporate Takeovers: Causes and Consequences, Alan J. Auerbach,
ed. Chicago: The University of Chicago Press, pp.
157-189.
Board of Governors of the Federal Reserve System.
1990. Flow of Fun& Accounts - Financial Assets and
Liabilities, Year-End 1965-1989. September.
Bagwell, Laurie Simon and Johh B Shoven. 1988.
"Share Repurchases and Acquisitions: An Analysis
of Which Firms Participate."
In Corporate Takeovers: Causes and Consequences, Alan J. Auerbach,
ed. Chicago: The University of Chicago Press, pp.
191-220.
Bolster, Paul J., Lawrence B. Lindsey, and Andrew
W. Mitrusi. 1989. "Tax-Induced Trading: The Effect of the 1986 Tax Reform Act on Stock Market
Activity." Journal of Finance 44, June, pp. 327-344.
Appendix:
Data
Sources
and
Campbell, Tim S. 1982. Financial Institutions, Markets, and Economic Activity. New York: McGrawHill Book Company.
Constantinides, George M. 1983. "Capital Market
Equilibrium with Personal Tax." Econonwtrica 51,
pp.611-636.
Dyl, Edward A. 1977. "Capital Gains Taxation and
the Year-W *mk, UI*djkUvior."
Journal of
Finance 32, March, pp. 165-175.
Ferris, Stephen P., Robert A. Haugen, and Aml K.
MakIWa. 1988. 'Tredicting Contemporary Volume
with Historic Volume at Differential Price Levels:
Evidence Supporting the Disposition Effect." Journal of Finance 43, July, pp. 677-697,
Henderson, Yolanda K. 1989. "Capital Gains Rates
and Revenues," New England Economic Review,
January/February,
pp. 3-20.
Karpoff, Jonathan M. 1986. "A Theory of Trading
Volume." Journal of Finance 41, December, pp.
1069-1087.
Kiefer, Donald W. 1990. "Lock-In Effect Within a
Simple Model of Corporate Stock Trading." Nauonal Tax Journal 43, March, pp. 75-94.
Lakoiushek, Joad and Seymour Smidt. 1986. "Volume for Winners and Losers: Taxation and Other
Motives for Stock 'rrading," Journal of Finance 41,
September, pp. 951-974.
Poterba, James M. 1987. "How Burdensome Are Capital Game Taxes? Evidence fiom the Umted States,"
Journal of Public Economics 33, July, pp. 157-172.
Shefrin, Hersh and Meir Statman. 1985. "The Disposition to Sell Winners Too Early and Ride Losers
Too Long: Theory and Evidence." Journal of Finance 40, July, pp. 777-790.
Slemrod, Joel. 1982. "Stock Transactions Volume and
the 1978 Capital Gains Tax Reduction." Public Finance Quarterly 10, January, pp. 3-16.
Toder, Eric. 1988. "Revenue Effects of Capital Gains
Taxes: Recent Time Series Evidence." Presented at
the American Economic Association Annual Meetings.
U.S. Congressional Budget Office. 1988. How Capital
Gains Tax Rates Affect Revenues: The Historical
Evidence. Washington,
D.C.: U.S. Government
Printing Office. March.
U.S. Treasury Department. 1985. Report to Congress
on the Capital Gains Tax Reductions of 1978. Washington, D.C.: U.S. Government Printing Office.
September.
Varian, Hal R. 1987. "Differences of Opinion in Financial Markets." J. Ira Harris Center for the Study
of Corporate Finance Working Paper No. 88-03. Ann
Arbor, Mich.. University of Michigan.
Methodology
Table 1
Average Daily Volume Sources:
NYSE New York Stock Exchange.
OTC
National Association of Securities
AMEX American Stock Exchange.
423
Dealers.
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
NATIONALTAX JOURNAL
424
[Vol. YLlll
AbsolutePercent Changein StockPrice Index
NYSE Absolute percentage change in the S&P Composite Index month-end to month-end.
Source: Board of Governors of the Federal Reserve System FAME data base and author',s calculations.
Absolute percentage change in the NASDAQ Composite Index month-end to monthOTC
end. Source: National Association of Securities Dealers, Inc., Fact Book 1989; Board
of Governors of the Federal Reserve System FAME data base; and author's calculations.
AMEX Absolute percentage change in the AMEX Market Value Index month-end to monthend. Source: American Stock Exchange, Marketing Research and Sales Support Department, "Amex Market Value Index, Month-end Closing Values, January 1969-September 1989"; Board of Gove@nors of the Federal Reserve System FAME data base;
and author's calculations.
Intra-Month Stock Price Volatility
Interquartile range of the daily percent change in closing values of the S&P 500 during the
month. Source: Board of Governors of the Federal Reserve System FAME data base and author's
calculations.
Absolute Monthly Change in 6-month interest rate, inflation rate and unemployment
Data Resources, Inc. and author's calculations.
rate. Source:
APPENDIX TABLE 1
NEW YORK STOCK EXCHANGE AVERAGE DAILY SALES VOLUME AND TURNOVER RATE, 1968-199o
Sales VoluMe (mils. of shares)
Total
Excl. Large Blocks
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
Iggoa
Jan.
Jun.
Sep.
1987
- May 1981
- Dec. 1981
- Dec. 1986
excl. Oct.
Turnover Rate (percent)
Total
Excl. Large Blocks
13.1
11.4
11.6
15.4
16.5
16.1
13.9
18.2
21.4
21.1
28.4
32.2
45.0
46.8
64.9
85.4
91.2
109.1
141.3
188.0
161.6
165.2
155.8
11.8
9.7
9.8
12.7
13.5
13.3
11.7
15.1
17.5
16.4
21.9
23.6
31.6
31.9
38.3
46.5
45.8
52.6
70.8
91.4
73.6
80.8
78.0
.106
.080
.074
.093
.091
.081
.065
.083
.092
.084
.106
.112
.142
.130
.167
.203
.192
.215
.255
.287
.218
.210
.185
.096
.069
.063
.077
.074
.066
.055
.069
.075
.065
.082
.082
.100
.089
.098
.111
.096
.104
.128
.140
.099
.103
.093
48.1
45.9
146.3
180.0
33.5
10.8
72.0
87.8
.140
.123
.251
.277
.098
.083
.124
.135
a First six months.
Source: See Appendix text.
National Tax Journal, Vol. 43, no. 4,
(December, 1990), pp. 411-25
No. 41
CAPITAL GAINS TAX RATES AND STOCK MARKET VOLUME
425
Table 2
Turnover rates = 100 x average daily volume/shares
month.
outstanding
at the end of the previous
Sources:
NYSE New York Stock Exchange and author's calculations.
National Association of@*g@curitiesDealers and authof s calculations.
OTC
AMEX American Stock Exchange and author's calculations.
Turnover rates less large blocks = 100 x (average daily volume - volume of large blocks)/shares
outstanding.
NYSE Source: New York Stock Exchange.
OTC
OTC large block data do not exist prior to 1983. However, large block trades were
msigmficant before that time. The data are smoothed to zero from 1983:1 back to 1982:2
in order to avoid a one month jump.
AMEX Data unavailable,
Absolute Percent Change in Stock Price Index
NYSE Absolute percent change in NYSE Common Stock Composite Index, month-end to monthend. Source: New York Stock Exchange, Business Research Department, "Common
Stock Indexes (1939-1988)," Board of Governors of the Federal Reserve System FAME
data base, and author's calculations.
OTC same as Table 1 above.
AMEX same as Table 1 above.
Intra-Month Stock Price Volatility
NYSE Monthly interquartile ranges for the NYSE Composite Index were calculated.
Source: Board of Governors of the Federal Reserve System FAME data base and author's calculations.
AMEX Monthly interquartile ranges for the NYSE Composite Index were calculated. Source:
Board of Governors of the Federal Reserve System FAME data base and author's calculations.
OTC Monthly interquartile ranges for the NASDAQ Composite Index were calculated. Source:
Board of Governors of the Federal Reserve System FAME data base and author's calculations.
Unexpected Change in the Inflation and Unemployment
Rates
The unexpected change was calculated as the absolute value of the difference between the actual
value of the variable and a forecast made from an ARIMA model. See Endnote 6.
NBER Forecasts
Two specifications were tried for the degree of economic disagreement for each variable. One
was the standard deviation divided by the mean, the other was the interquartile range divided
by the median.
Retail Trade Data (Figure 1)
The Securities Industry Association publishes monthly data on retail, member, and institutional trading for the NYSE and AMEX in its Investor Activity Report. Data are published on
institutional trading for the OTC. The data are available back to mid- 1985. In addition, a series
of adjusted estimates (recently revised) is available for the NYSE back to January 1987. This
series subtracts all dividend capture and three-fourths of program trading from the retail category and more accurately reflects trading by individual investors.